Final Results for the Solenoid Magnetic Field Map CERN mapping project team Martin Aleksa, Felix Bergsma, Laurent Chevalier, Pierre-Ange Giudici, Antoine Kehrli, Marcello Losasso, Xavier Pons, Heidi Sandaker Map fitting by John Hart (RAL) Paul S Miyagawa, Steve Snow (Manchester)

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20072/20 Outline Overview of mapping campaign Corrections to data Geometrical fit results Geometrical + Maxwell fit results Systematic errors Conclusions

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20073/20 Overview of the task Mapping 6m long x 2m diameter cylindrical volume 2 Tesla (20000 Gauss) at Z=0, dropping to 0.8 T at Z=3m Defines momentum scale of all ID tracks Require track sagitta error due to field uncertainty < 0.05% –Ensures that error in momentum due to field is less than error due to tracker alignment at p T of 40 GeV

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20074/20 Cards each hold 3 orthogonal sensors 4 fixed NMR probes at Z=0 4 arms in windmill. Each arm equipped with 12 Hall cards Pneumatic motors with optical encoders. Move and measure in Z and φ. The field mapping machine

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20075/20 Data sets recorded Data taken at four different solenoid currents –Nominal current (7730 A) gives 2 T at centre –Low current (5000 A) gives 1.3 T and is used with low-field probe calibration Fine phi scans used to measure the (tiny) perturbation of the field by the mapping machine Total data collected ~0.75M –Statistical errors will be negligible Date in August Current (A) Number of φ steps Number of Z steps 2nd-3rd rd th th

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20076/20 Corrections to data Geometrical effects Plenty of survey data taken before and after mapping campaign –Positions of individual Hall sensors can be determined to ~0.2 mm accuracy Mapping machine skew recorded in data Carriage tilts determined from data Probe calibrations Response of Hall sensors calibrated as function of field strength, field orientation and temperature using test stands at CERN and Grenoble –Low-field calibration (up to 1.4 T) has expected accuracy of ~2 G, 2 mrad –High-field calibration (up to 2.5 T) has expected accuracy of ~10 G, 2 mrad NMR probes intrinsically accurate to 0.1 G Absolute scale of high-field Hall calibration improved using low-field Hall calibration and NMR values Relative Hall probe normalisations and alignments determined from data Other effects Effects of magnetic components of mapping machine corrected using magnetic dipoles

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20077/20 Probe normalisation and alignment Exploited the mathematics of Maxwell’s equations to determine relative probe normalisations and alignments B Z normalisation: –Uses the fact that each probe scans the field on the surface of a cylinder –B Z at centre determined for each probe –All probes were then normalised to the average of these values Probe alignment: –Uses curl B = 0 and –Integrate –Tilt angles A ij of probe were determined from a least squares fit –The third alignment angle comes from div B = 0

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20078/20 Carriage tilts Another analysis which exploited mathematics of Maxwell B x and B y on the z-axis evaluated from average over φ for probes near centre of solenoid Plots of B x,B y versus Z of carriage show evidence that entire carriage is tilting Degree of tilt can be calculated by integrating to find expected B x,B y value Jagged structure of tilts suggest that machine is going over bumps on the rail of ~0.1 mm

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July 20079/20 Absolute Hall scale Absolute scale of high-field Hall calibration (10 G) is greatest uncertainty –Can be improved using low-field Hall calibration (2 G) and NMR value (0.1 G) Low-field Hall values and NMR values are equal for 5000 A data –Low-field Hall values are considered accurate in low-field region Discrepancy between low- and high-field Hall values in low-field region Discrepancy between high-field Hall values (derived from field fits) and NMR value in high-field region –This discrepancy lines up with the discrepancy from low-field region Alternative high-field Hall values from extrapolation give estimate of systematic error

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Fit quality measures We fit the map data to field models which obey Maxwell’s equations –The volume covered has no currents and has effects of magnetic materials removed –Maxwell’s equations become Our fit uses Minuit to minimise Our aim is to know the track sagitta, which is proportional to (c r and c z are direction cosines) Our fit quality is defined to be δS/S where

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Geometrical fit 96% of the field is directly due to the solenoid current –We use a detailed model of the conductor geometry and integrate Biot-Savart law using the known current –7 free parameters Scale factor and aspect ratio (length/diameter) of conductor model Position and orientation of conductor model relative to IWV 4% of field is due to magnetised iron (TileCal, girders, shielding discs etc) –Parametrised using 4 free parameters of Fourier-Bessel series with length scale=2.5m Conductor geometry determined by surveys of solenoid as built except weld thickness, which was determined from data as 1.9×pitch

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Results from geometrical fit I B Z B R B φ 20 G 40 G

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Results from geometrical fit II Residuals + sagitta error calculated for all data samples –7730a used for our final fit results at nominal 2 T field Sagitta error rms is within our 5×10 -4 target for all samples MapB Z (G)B R (G)B φ (G)δS/S (×10 -4 ) rms extreme rms extreme rms extreme rms extreme h a b

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Results from geometrical fit III Fit parameters tabulated for all samples All parameters consistent with expected results –Expected offsets for solenoid centre: x = -0.3 ± 0.4 mm, y = -2.2 ± 0.4 mm, z = -0.1 ± 2.3 mm –Z and R scale factors expected to be 1 MapOffsets (mm)Angles (mrad)Scale factors% iron at orig xyzAxAx AyAy ZR h a b

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Full fit (geometrical + Maxwell) A few features remain in the residuals from the geometrical fit –Ripples for |Z|<2m believed to be due to variations in the coil winding density –Bigger features at |Z|>2m believed to arise from the coil cross-section not being perfectly circular These effects are more pronounced at the ends of the solenoid These features cannot be determined accurately enough to be included in the geometrical model –However, they are real fields which should obey Maxwell’s equations We apply the general Maxwell fit to the residuals to account for these features

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 General Maxwell fit General fit able to describe any field obeying Maxwell’s equations. Uses only the field measurements on the surface of a bounding cylinder, including the ends. Parameterisation proceeds in three stages: 1.B z on the cylindrical surface is fitted as Fourier series, giving terms with φ variation of form cos(nφ+α), with radial variation I n (κr) (modified Bessel function). 2.B z meas – B z (1) on the cylinder ends is fitted as a series of Bessel functions, J n (λ j r) where the λ j are chosen so the terms vanish for r = r cyl. The z-dependence is of form cosh(μz) or sinh(μz). 3.The multipole terms are calculated from the measurements of B r on the cylindrical surface, averaged over z, after subtraction of the contribution to B r from the terms above. (The only relevant terms in B z are those that are odd in z.)

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Results from full fit I B Z B R B φ 20 G 40 G

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Results from full fit II Residuals of all probes reduced significantly –Recall that Maxwell fit is made using outermost probes only –Fact that the fit matches inner probes as well shows strong evidence that the difference between data and geometrical model is a real field Fit quality δS/S also improved at high η MapB Z (G)B R (G)B φ (G)δS/S (×10 -4 ) rms extreme rms extreme rms extreme rms extreme h a b

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Systematic errors Uncertainty in overall scale –Spread in Hall-NMR differences over 4 NMR probes –Weld thickness, which influences the Hall-NMR comparison –Overall scale error 2.1×10 -4 Uncertainty in shape of field –Considered several “reasonable” changes to fit model –Dominant factor is 0.2 mrad uncertainty in orientation of the rotation axes of the mapping machine arms relative to IWV coordinates –Overall shape error 5.9×10 -4 Total uncertainty 6.3×10 -4 –Dominated by scale error at low η, shape error at high η

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Conclusions The solenoid field mapping team recorded lots of high quality data during a successful field mapping campaign All possible corrections from surveys, probe calibrations and probe alignments have been applied to the data We have determined a fit function satisfying Maxwell’s equations which matches each component of the data to within 4 Gauss rms Using this fit, the relative sagitta error is 6.3×10 -4 At high rapidity, the systematic errors are dominated by a 0.2 mrad uncertainty in the direction of the field axis relative to the IWV coordinate system

Backup slides

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Surveys Survey of mapping machine in Building 164 –Radial positions of Hall cards –Z separation between arms –Z thickness of arms Survey in situ before and after mapping –Rotation centre and axis of each arm –Position of Z encoder zero –Positions of NMR probes Survey of ID rails –Gradient wrt Inner Warm Vessel coordinates Survey of a sample of 9 Hall cards –Offsets of BZ, BR, Bf sensors from nominal survey point on card Sensor positions known with typical accuracy of 0.2 mm

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Probe calibrations Hall sensors Response measured at several field strengths, temperatures and orientations (θ,φ) Hall voltage decomposed as spherical harmonics for (θ,φ) and Chebyshev polynomials for |B|,T Low-field calibration (up to 1.4 T): expected accuracy ~2 G, 2 mrad High-field calibration (up to 2.5 T): expected accuracy ~10 G, 2 mrad NMR probes No additional calibration needed (done by whoever measured Gp = MHz/T) Compare proton resonance frequency with reference oscillator Intrinsically accurate to 0.1 G

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Magnetic machine components

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Grid and test points Extra grid points near special features (ends of solenoid, welds, return conductor)

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Linear interpolation Linear search to find nearest grid points Linear interpolation in each direction Field components treated in cylindrical coordinates

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Linear interpolation

Paul S MiyagawaATLAS Overview Week, Glasgow, 12 July /20 Linear interpolation